Why does MCP cause token bloat?

Every MCP tool connected injects its full JSON schema into context on every turn. 30 tools over 15 turns burns 96% more tokens just on schema, before any reasoning happens.

How bad is token usage with 120 MCP tools?

120 tools over 25 turns burns 362k tokens on schema descriptions alone before the model does anything useful. That is the number mcp2cli measured and publicized on HN.

What does mcp2cli do?

mcp2cli converts any MCP server or OpenAPI spec into a CLI at runtime with no code generation. Agents discover tools on demand instead of loading all schemas upfront.

How much does mcp2cli reduce token usage?

96% reduction for 30 tools over 15 turns, 99% for 120 tools over 25 turns. Measured with cl100k_base.

Is mcp2cli the only solution to MCP token bloat?

mcp2cli (2.2k stars) is the main CLI approach. Server-side MCP gateways and Bifrost address the same problem from the infrastructure layer. No dominant winner yet.

← Back to dashboard

clawsmith.com/signal/mcp-token-context-bloat-cli-discovery

⚠ IssueUnderserveddev_tool_cliLive

Connecting 120 MCP tools burns 362k tokens per session just on schema injection before the model does anything useful

Every MCP tool connected to an LLM agent injects its full JSON schema into the context window on every turn. 120 tools over 25 turns burns 362k tokens on schema descriptions alone. mcp2cli launched on HN (146 pts, 73 comments) solving this by turning any MCP server or OpenAPI spec into a CLI with on-demand tool discovery, cutting token use by 96-99%. The tool has 2.2k GitHub stars.

Product Idea from this Signal

A CLI tool that converts any MCP server into token-efficient CLI commands at runtime

5.1k ▲

MCP tool schema injection burns 50,000-67,000 tokens before an agent types a single prompt, with GitHub MCP alone costing 55,000 tokens across 93 tool definitions. Developers building AI agents pay 4-32x more per operation than they would using CLI equivalents, but manually rewriting each MCP server as a CLI is impractical at scale. This tool converts any MCP server or OpenAPI spec into lightweight CLI commands at runtime, letting agents execute the same operations at 96-99% lower token cost without losing any capability.

mcpai-agentstoken-efficiencyclideveloper-toolsllm-costopenapi

Competitive220 leadsView Opportunity →

Score Breakdown

GitHub

2,295

219

Social Proof 2 sources

knowsuchagency/mcp2cli – Turn any MCP server or OpenAPI spec into a CLI

gh:knowsuchagency · 3/9/2026

2,295 HN

Show HN: Mcp2cli – One CLI for every API, 96-99% fewer tokens than native MCP

knowsuchagency · 3/9/2026

219

Existing Solutions 2 competitors

mcp2cli2.2k GitHub stars, 146 HN points

Turns any MCP server or OpenAPI spec into a CLI at runtime

BifrostEarly OSS, server-side only

Open-source LLM gateway reducing token usage by 50% in multi-tool workflows

Gap Assessment

UnderservedExisting solutions leave gaps

mcp2cli (2.2k stars, HN front page) is the main early solution; Bifrost and MCP gateways exist but are server-side, not CLI-native; category is growing fast.

Frequently Asked Questions

Virality Score

2,514

across 0 platforms

Details

Signalissue

Ecosystemdev_tool_cli

Sources2

Platforms0

Updatedunknown

Trend→ stable

Top ideas

All ideas →

0A CLI tool that wraps stateful MCP servers and externalizes their session state so they run behind standard round-robin load balancers without sticky routing 0A CLI tool that tails and greps logs across multiple remote hosts in a single TUI without centralized infrastructure 0A browser extension that removes Gemini and all Google AI injections from Chrome across Search, Gmail, Docs, and Drive in one toggle